Control system and control method of intelligent PON home gateway equipment

Abstract

The invention relates to a control system and control method for intelligent PON home gateway equipment, and the method comprises the steps: collecting the operation data of gateway equipment and the application behavior data of terminal equipment in real time, and carrying out the fusion analysis based on a deep neural network and a reinforcement learning algorithm; constructing a decision network comprising a plurality of sub decision units, and generating a corresponding control instruction based on gateway equipment operation data and terminal equipment application behavior data collected in real time; according to the real-time load condition of each functional module of the home gateway equipment, intelligently distributing an instruction execution priority, transmitting an instruction to the corresponding functional module, and driving the functional module to execute instruction operation; and the execution effect of the control instruction is monitored from multiple dimensions of performance indexes, user experience and energy consumption, the executed equipment operation data and the evaluation result are fed back to the data analysis module for analysis, and decision dynamic adjustment is performed based on the analysis result. Fine management and optimization of the home gateway equipment are realized, and network performance and user experience are effectively improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of home gateway intelligent control, and in particular relates to a control system and a control method for an intelligent PON home gateway device. Background Art

[0002] With the rapid development of Internet technology and the widespread adoption of smart home devices, the performance and functionality of smart PON home gateways, as the core hub connecting home networks with external networks, directly impact users' network experience and the implementation of smart home applications. Existing smart PON home gateway device control systems and control methods have many limitations in practical applications, making it difficult to meet the growing network demands and intelligent management requirements.

[0003] In terms of data collection and analysis, traditional home gateway device control systems have a relatively simple data collection method, only able to obtain basic operating parameters of the gateway device itself, such as network connection status and data flow, and unable to comprehensively collect application behavior data from various terminal devices on the home network. Moreover, there is a lack of effective integrated analysis methods for the collected data, and data processing remains at the simple monitoring and statistical level, making it difficult to deeply explore the data value, accurately determine the operating status of devices, and predict network traffic trends. As a result, the system cannot be optimized and adjusted in advance, affecting network stability and user experience.

[0004] In the decision-making and control stages, existing control system decision-making processes are often based on fixed, preset rules, lacking flexibility and adaptability. When the network environment changes or complex failure scenarios arise, fixed rules make it difficult to make optimal decisions. For example, during network congestion, traditional systems may be unable to dynamically adjust bandwidth allocation based on the priorities and real-time needs of different applications, affecting the service quality of critical applications (such as video conferencing and online gaming). Furthermore, the instruction execution process lacks dynamic awareness and intelligent scheduling of device load conditions, making it impossible to properly allocate instruction execution priorities. This can easily lead to resource waste caused by overloading some devices while leaving others idle, reducing the overall operational efficiency of the system.

[0005] In terms of feedback and optimization, traditional control systems lack robust feedback mechanisms, primarily focusing on simple feedback on equipment operating status and lacking comprehensive assessments of multiple dimensions, such as user experience and energy consumption. Furthermore, the processing and utilization of feedback information is inefficient, making it difficult to effectively support continuous system optimization. Furthermore, existing systems lack effective traceability of control command execution, making it difficult to accurately identify the cause of problems, hindering system maintenance and improvement.

[0006] With the continuous increase in the number of devices in home networks, the increasing complexity of application scenarios, and the continuous improvement of users' requirements for network quality and intelligent management, there is an urgent need for a more intelligent, efficient, and comprehensive control system and control method for intelligent PON home gateway devices to achieve refined management and optimization of home gateway devices, improve network performance and user experience, and meet the needs of future home network development. Summary of the Invention

[0007] The purpose of the present invention is to provide a control system and control method for an intelligent PON home gateway device, so as to achieve refined management and optimization of the home gateway device, improve network performance and user experience, and meet the needs of future home network development.

[0008] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows:

[0009] In a first aspect, a control system for an intelligent PON home gateway device is provided, comprising a data acquisition module, a data analysis module, a dynamic collaborative decision-making module, an adaptive execution module, and a feedback evaluation module, wherein the data acquisition module is connected to the data analysis module, the data analysis module is connected to the dynamic collaborative decision-making module, the dynamic collaborative decision-making module is connected to the adaptive execution module, the adaptive execution module is connected to the feedback evaluation module, and the feedback evaluation module is connected to the dynamic collaborative decision-making module;

[0010] The data acquisition module is used to collect the operating data of each module of the PON home gateway device and the application behavior data of each terminal device in the home network in real time;

[0011] The data analysis module is used to perform a fusion analysis of the collected operation data of the gateway device and the application behavior data of the terminal device based on a deep neural network and a reinforcement learning algorithm;

[0012] The dynamic collaborative decision-making module is used to build a decision-making network including multiple sub-decision-making units, and generate corresponding control instructions based on the real-time collected gateway device operation data and terminal device application behavior data;

[0013] The adaptive execution module is used to receive the control instruction and intelligently assign the instruction execution priority according to the real-time load of each functional module of the home gateway device, transmit the instruction to the corresponding functional module, and drive the functional module to execute the instruction operation;

[0014] The feedback evaluation module is used to monitor the execution effect of control instructions from multiple dimensions such as performance indicators, user experience, and energy consumption, and uses blockchain technology to store the execution process. The post-execution equipment operation data and evaluation results are fed back to the data analysis module for analysis. The dynamic collaborative decision-making module dynamically adjusts decisions based on the analysis results.

[0015] Preferably, the data acquisition module includes an edge computing node and a distributed sensor array, the distributed sensor array is deployed at designated key locations of the home gateway device, the edge computing node establishes a connection with each sensor via a communication protocol such as I2C or SPI, the distributed sensor array converts the collected physical signal into an electrical signal and transmits it to the corresponding edge computing node, and the node's built-in analog-to-digital converter converts the electrical signal into digital data;

[0016] At the same time, the edge node communicates with the terminal devices in the home network through the network interface, using the simple network management protocol or the API interface customized by the device manufacturer to collect the application behavior data of the terminal devices. The collected data is preliminarily preprocessed at the edge computing node and packaged according to the preset data structure, and transmitted to the data analysis module through the high-speed data bus.

[0017] Preferably, the data analysis module includes a deep neural network processing unit and a reinforcement learning algorithm execution unit. After receiving various types of data obtained by each edge node, the deep neural network processing unit performs feature extraction on each type of data to obtain high-dimensional data features, and reduces the dimension of the high-dimensional data features through the principal component analysis algorithm to obtain a low-dimensional feature vector, and transmits the low-dimensional feature vector to the reinforcement learning algorithm execution unit.

[0018] Preferably, the reinforcement learning algorithm execution unit constructs the environment state based on the low-dimensional feature vector, calculates the action value of each possible action, i.e., the Q value, based on the deep neural network processing unit, and the deep neural network processing unit selects actions based on the ∈-greedy strategy, i.e., randomly selects actions with a certain probability ∈, explores new environment states, and finds possible better strategies; selects the action that the current strategy network believes has the largest Q value with a probability of 1-∈. In the early stage of the algorithm, ∈ takes a larger value to encourage more exploratory behaviors; as the training progresses, ∈ gradually decreases, and more use is made of the learned strategies.

[0019] Preferably, the execution unit of the reinforcement learning algorithm execution unit executes action A t After that, the state of the home gateway device will transition to the new state S t+1 The execution unit quantitatively evaluates the effect of the action based on the pre-designed reward mechanism and calculates the corresponding immediate reward R t+1The reward function is designed based on the optimization goal specified by the anchoring system. The specified optimization goal includes network performance optimization or device stability optimization. In the reward calculation process, the reward attenuation factor γ is set, γ∈(0,1) to balance short-term rewards and long-term rewards, so that the algorithm pays more attention to long-term benefits. Gt=R t+1 +γR t+2 +γ2R t+3 +…;

[0020] Execute action A t After that, the actual operation status of the home gateway device changes from S t Transfer to S t+1 This process is determined by the actual operation mechanism of the home gateway device, including the action of adjusting the wireless channel parameters, which will change the signal strength and channel occupancy of the wireless communication module, thereby affecting the performance of the entire home network. These changes are reflected in the new state S t+1 In the process of executing the unit to record the state transition, that is, (S t ,A t ,R t+1 ,S t+1 ), and stores the data in the experience replay buffer.

[0021] Preferably, the multiple sub-decision-making units of the dynamic collaborative decision-making module include an optical signal processing sub-decision-making unit, a CPU load management sub-decision-making unit, and a network traffic scheduling sub-decision-making unit. After receiving the analysis results of the data analysis module, the dynamic collaborative decision-making module first determines the activated sub-decision-making unit through the attention mechanism according to the current network scenario requirements. When the optical signal strength of the PON interface module is detected to be abnormal, the optical signal processing sub-decision-making unit is activated; if it is predicted that the network traffic will increase significantly, the network traffic scheduling sub-decision-making unit is activated at the same time;

[0022] The sub-decision-making unit makes local decisions based on the preset control strategy and the information received. The preset control strategy is optimized and adjusted according to the historical decision-making results and changes in the network environment through the reinforcement learning algorithm. When it is judged that the optical signal strength is lower than the threshold, the Q-Learning algorithm is used to select a joint strategy of adjusting the optical module transmission power and switching to the backup optical path based on the optical signal recovery situation under different adjustment strategies in the historical data, and generate corresponding control instructions. The decision results of multiple sub-decision-making units are integrated through a multi-objective optimization algorithm to balance the conflicts between different decisions, minimize energy consumption while ensuring network stability, and finally generate the globally optimal control instructions and transmit them to the adaptive instruction execution module.

[0023] Preferably, the adaptive execution module includes a software-defined network architecture, which includes a controller and a forwarder. After receiving the control instruction generated by the dynamic collaborative decision-making module, the controller first parses the instruction content to determine the functional modules and operation types involved in the home gateway device. Then, through the real-time communication connection established with each functional module, the controller obtains the current load status of the functional module, such as the remaining CPU resources of the data processing module and the channel occupancy rate of the wireless communication module.

[0024] Based on load conditions, the controller uses a load balancing algorithm and a minimum connection count algorithm to prioritize control commands. Urgent commands with significant system impact, such as those addressing network outages, are given a higher priority. Less urgent optimization commands, such as those adjusting wireless channel parameters, are given a lower priority. After assigning priorities, the controller accurately transmits the commands to the forwarder using SDN communication protocols such as OpenFlow. Upon receiving the commands, the functional modules execute the actions specified in the instructions, such as shutting down non-critical application processes in the data processing module or switching wireless channels in the wireless communication module. The results are then fed back to the controller.

[0025] Preferably, the feedback evaluation module monitors and evaluates the execution effect of the control instruction from multiple dimensions of performance indicators, user experience, and energy consumption;

[0026] In terms of performance indicator monitoring, probes deployed at key network nodes collect data such as network bandwidth utilization, latency, and packet loss rate in real time. The built-in performance monitoring module of the device is used to obtain indicators such as device processing capacity and response time.

[0027] In user experience evaluation, a combination of active detection and passive monitoring is used. Test data packets are sent to terminal devices in the home network to assess network quality. User feedback on network usage, such as video freezes and game delays, is passively collected through apps or web pages. For energy consumption monitoring, the power consumption of home gateway devices is recorded in real time through devices such as smart meters.

[0028] Integrating blockchain technology, key data from the control command execution process, such as command content, execution time, and the responses of each functional module, is packaged according to the blockchain's block structure and a unique block identifier is generated using a hash algorithm. Each block is linked in sequence to form an unalterable chain of evidence of the execution process. The module integrates and analyzes the collected post-execution device operation data with the evaluation results, generating a feedback report that is transmitted to the data analysis module, providing a basis for subsequent data analysis and decision adjustments. It also supports users or administrators in conducting secure and reliable retrospective audits of the control command execution process.

[0029] In a second aspect, a control method for an intelligent PON home gateway device is provided, comprising the following steps:

[0030] S1: Real-time collection of operating data of each module of the PON home gateway device and application behavior data of each terminal device in the home network;

[0031] S2: Based on deep neural networks and reinforcement learning algorithms, it integrates and analyzes the collected operation data of gateway devices and application behavior data of terminal devices;

[0032] S3: Build a decision-making network consisting of multiple sub-decision-making units to generate corresponding control instructions based on the real-time collected gateway device operation data and terminal device application behavior data;

[0033] S4: Based on the received control instruction and according to the real-time load of each functional module of the home gateway device, intelligently assigning the instruction execution priority, transmitting the instruction to the corresponding functional module, and driving the functional module to execute the instruction operation;

[0034] S5: Monitor the execution effect of control instructions from multiple dimensions such as performance indicators, user experience, and energy consumption, use blockchain technology to store the execution process, and feed back the equipment operation data and evaluation results after execution to the data analysis module for analysis. The dynamic collaborative decision-making module dynamically adjusts decisions based on the analysis results.

[0035] The beneficial effects of the present invention include:

[0036] The control system and control method for intelligent PON home gateway devices provided by the present invention collect real-time gateway device operation data and terminal device application behavior data, and perform fusion analysis based on deep neural networks and reinforcement learning algorithms. A decision network consisting of multiple sub-decision-making units is constructed to generate corresponding control instructions based on the real-time collected gateway device operation data and terminal device application behavior data. The control network intelligently assigns instruction execution priorities based on the real-time load status of each functional module of the home gateway device, transmits instructions to the corresponding functional modules, and drives the functional modules to execute the instruction operations. The control system monitors the execution effect of control instructions from multiple dimensions such as performance indicators, user experience, and energy consumption. The executed device operation data and evaluation results are fed back to the data analysis module for analysis, and dynamic decision adjustments are made based on the analysis results. This achieves refined management and optimization of home gateway devices, effectively improving network performance and user experience.

[0037] First, the data acquisition module uses edge computing technology combined with a distributed sensor array to collect and fuse multi-source heterogeneous data. This not only collects operational data from various modules of home gateway devices, but also obtains application behavior data from various terminal devices in the home network, greatly enriching the data sources. The data analysis module uses deep neural networks and reinforcement learning algorithms to fuse and analyze multi-source data. A dynamic threshold adaptive algorithm and trend prediction model automatically adjust the threshold for determining device operating status, effectively improving threshold accuracy. It also accurately predicts network traffic trends, enabling precise judgment of device operating status and early optimization of network performance. This significantly enhances network stability and reliability, and improves the user's network experience.

[0038] Secondly, by building a dynamic collaborative decision-making module consisting of multiple sub-decision-making units, the corresponding sub-decision-making units can be dynamically activated or dormant according to different network scenarios, reducing system decision-making power consumption. When faced with complex network environments or fault scenarios, more optimized control instructions can be generated based on abnormal signals, network traffic trends, and dynamically updated preset control strategies.

[0039] Finally, the execution of control commands is monitored across multiple dimensions, including performance indicators, user experience, and energy consumption, and evidence of the execution process is stored using blockchain technology. This multi-dimensional, comprehensive evaluation and tamper-proof evidence storage mechanism not only provides a reliable basis for traceability of control command execution, ensuring system security and auditability, but also facilitates a comprehensive and accurate assessment of system performance, providing richer and more effective feedback to the data analysis module, supporting continuous system optimization, and further enhancing the intelligent management level and overall performance of home gateway devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 Schematic diagram of the control system architecture of the intelligent PON home gateway device of the present invention.

[0041] Figure 2 It is a flow chart of a control method of an intelligent PON home gateway device of the present invention. DETAILED DESCRIPTION

[0042] The following is combined with Figures 1 and 2 The present invention is described in further detail:

[0043] Example 1

[0044] See attached Figure 1As shown, a control system of an intelligent PON home gateway device includes a data acquisition module, a data analysis module, a dynamic collaborative decision-making module, an adaptive execution module, and a feedback evaluation module. The data acquisition module is connected to the data analysis module, the data analysis module is connected to the dynamic collaborative decision-making module, the dynamic collaborative decision-making module is connected to the adaptive execution module, the adaptive execution module is connected to the feedback evaluation module, and the feedback evaluation module is connected to the dynamic collaborative decision-making module.

[0045] The data acquisition module is used to collect real-time operational data from each module of the PON home gateway device and application behavior data from each terminal device in the home network. The data analysis module is used to perform a fusion analysis of the collected gateway device operational data and terminal device application behavior data based on a deep neural network and reinforcement learning algorithm. The dynamic collaborative decision-making module is used to construct a decision-making network consisting of multiple sub-decision-making units to generate corresponding control instructions based on the real-time collected gateway device operational data and terminal device application behavior data. The adaptive execution module is used to receive the control instructions and intelligently assign instruction execution priorities based on the real-time load of each functional module of the home gateway device. The instructions are transmitted to the corresponding functional modules to drive the functional modules to execute the instruction operations. The feedback evaluation module is used to monitor the execution effect of the control instructions from multiple dimensions such as performance indicators, user experience, and energy consumption. The execution process is recorded using blockchain technology. The executed device operational data and evaluation results are fed back to the data analysis module for analysis. The dynamic collaborative decision-making module dynamically adjusts its decisions based on the analysis results.

[0046] Example 2

[0047] Based on Example 1, the data acquisition module includes an edge computing node and a distributed sensor array. The distributed sensor array is deployed at designated key locations of the home gateway device. The edge computing node establishes a connection with each sensor through communication protocols such as I2C and SPI. The distributed sensor array converts the collected physical signals into electrical signals and transmits them to the corresponding edge computing node. The analog-to-digital converter built into the node converts the electrical signals into digital data.

[0048] At the same time, the edge node communicates with the terminal devices in the home network through the network interface, using the simple network management protocol or the API interface customized by the device manufacturer to collect the application behavior data of the terminal devices. The collected data is preliminarily preprocessed at the edge computing node and packaged according to the preset data structure, and transmitted to the data analysis module through the high-speed data bus.

[0049] In this embodiment, the data analysis module includes a deep neural network processing unit and a reinforcement learning algorithm execution unit. After the deep neural network processing unit receives various types of data obtained by each edge node, it performs feature extraction on each type of data to obtain high-dimensional data features, and reduces the dimension of the high-dimensional data features through the principal component analysis algorithm to obtain a low-dimensional feature vector, and transmits the low-dimensional feature vector to the reinforcement learning algorithm execution unit.

[0050] The reinforcement learning algorithm execution unit constructs the environment state based on the low-dimensional feature vector, and calculates the action value of each possible action, i.e., the Q value, based on the deep neural network processing unit. The deep neural network processing unit selects actions based on the ∈-greedy strategy, i.e., randomly selects actions with a certain probability ∈, explores new environment states, and discovers potentially better strategies; selects the action that the current strategy network believes has the largest Q value with a probability of 1-∈. In the early stages of the algorithm, ∈ takes a larger value to encourage more exploratory behavior; as training progresses, ∈ gradually decreases, and more use is made of the learned strategies.

[0051] The execution unit of the reinforcement learning algorithm execution unit executes action A t After that, the state of the home gateway device will transition to the new state S t+1 The execution unit quantitatively evaluates the effect of the action based on the pre-designed reward mechanism and calculates the corresponding immediate reward R t+1 The reward function is designed based on the optimization goal specified by the anchoring system. The specified optimization goal includes network performance optimization or device stability optimization. In the reward calculation process, the reward attenuation factor γ is set, γ∈(0,1) to balance short-term rewards and long-term rewards, so that the algorithm pays more attention to long-term benefits. Gt=R t+1 +γR t+2 +γ2R t+3 +…;

[0052] Execute action A t After that, the actual operation status of the home gateway device changes from S t Transfer to S t+1 This process is determined by the actual operation mechanism of the home gateway device, including the action of adjusting the wireless channel parameters, which will change the signal strength and channel occupancy of the wireless communication module, thereby affecting the performance of the entire home network. These changes are reflected in the new state S t+1 In the process of executing the unit to record the state transition, that is, (S t ,A t ,R t+1 ,S t+1 ), and stores the data in the experience replay buffer.

[0053] Example 3

[0054] Based on Example 1 or Example 2, the multiple sub-decision units of the dynamic collaborative decision-making module include an optical signal processing sub-decision unit, a CPU load management sub-decision unit and a network traffic scheduling sub-decision unit. After receiving the analysis results of the data analysis module, the dynamic collaborative decision-making module first determines the activated sub-decision unit through the attention mechanism based on the current network scenario requirements. When the optical signal strength of the PON interface module is detected to be abnormal, the optical signal processing sub-decision unit is activated; if it is predicted that the network traffic will increase significantly, the network traffic scheduling sub-decision unit is activated at the same time.

[0055] The sub-decision-making unit makes local decisions based on the preset control strategy and the information received. The preset control strategy is optimized and adjusted according to the historical decision-making results and changes in the network environment through the reinforcement learning algorithm. When it is judged that the optical signal strength is lower than the threshold, the Q-Learning algorithm is used to select a joint strategy of adjusting the optical module transmission power and switching to the backup optical path based on the optical signal recovery situation under different adjustment strategies in the historical data, and generate corresponding control instructions. The decision results of multiple sub-decision-making units are integrated through a multi-objective optimization algorithm to balance the conflicts between different decisions, minimize energy consumption while ensuring network stability, and finally generate the globally optimal control instructions and transmit them to the adaptive instruction execution module.

[0056] In this embodiment, the adaptive execution module includes a software-defined network architecture, which includes a controller and a forwarder. After receiving the control instruction generated by the dynamic collaborative decision-making module, the controller first parses the instruction content to determine the functional modules and operation types involved in the home gateway device. Then, through real-time communication connections established with each functional module, it obtains the current load status of the functional module, such as the remaining CPU resources of the data processing module and the channel occupancy rate of the wireless communication module.

[0057] Based on load conditions, the controller uses a load balancing algorithm and a minimum connection count algorithm to prioritize control commands. Urgent commands with significant system impact, such as those addressing network outages, are given a higher priority. Less urgent optimization commands, such as those adjusting wireless channel parameters, are given a lower priority. After assigning priorities, the controller accurately transmits the commands to the forwarder using SDN communication protocols such as OpenFlow. Upon receiving the commands, the functional modules execute the actions specified in the instructions, such as shutting down non-critical application processes in the data processing module or switching wireless channels in the wireless communication module. The results are then fed back to the controller.

[0058] The feedback evaluation module monitors and evaluates the execution effect of the control instructions from multiple dimensions such as performance indicators, user experience, and energy consumption;

[0059] In terms of performance indicator monitoring, probes deployed at key network nodes collect data such as network bandwidth utilization, latency, and packet loss rate in real time. The built-in performance monitoring module of the device is used to obtain indicators such as device processing capacity and response time.

[0060] In user experience evaluation, a combination of active detection and passive monitoring is used. Test data packets are sent to terminal devices in the home network to assess network quality. User feedback on network usage, such as video freezes and game delays, is passively collected through apps or web pages. For energy consumption monitoring, the power consumption of home gateway devices is recorded in real time through devices such as smart meters.

[0061] Integrating blockchain technology, key data from the control command execution process, such as command content, execution time, and the responses of each functional module, is packaged according to the blockchain's block structure and a unique block identifier is generated using a hash algorithm. Each block is linked in sequence to form an unalterable chain of evidence of the execution process. The module integrates and analyzes the collected post-execution device operation data with the evaluation results, generating a feedback report that is transmitted to the data analysis module, providing a basis for subsequent data analysis and decision adjustments. It also supports users or administrators in conducting secure and reliable retrospective audits of the control command execution process.

[0062] A control method for an intelligent PON home gateway device, see Figure 2 , including the following steps:

[0063] S1: Real-time collection of operating data of each module of the PON home gateway device and application behavior data of each terminal device in the home network;

[0064] S2: Based on deep neural networks and reinforcement learning algorithms, it integrates and analyzes the collected operation data of gateway devices and application behavior data of terminal devices;

[0065] S3: Build a decision-making network consisting of multiple sub-decision-making units to generate corresponding control instructions based on the real-time collected gateway device operation data and terminal device application behavior data;

[0066] S4: Based on the received control instruction and according to the real-time load of each functional module of the home gateway device, intelligently assigning the instruction execution priority, transmitting the instruction to the corresponding functional module, and driving the functional module to execute the instruction operation;

[0067] S5: Monitor the execution effect of control instructions from multiple dimensions such as performance indicators, user experience, and energy consumption. Combined with blockchain technology, the execution process is recorded. The equipment operation data and evaluation results after execution are fed back to the data analysis module for analysis. The dynamic collaborative decision-making module makes dynamic adjustments based on the analysis results.

[0068] In summary, the control system and control method of the intelligent PON home gateway device provided by the present invention uses edge computing technology combined with a distributed sensor array to collect and fuse multi-source heterogeneous data through the data acquisition module. It can not only collect the operating data of each module of the home gateway device, but also obtain the application behavior data of each terminal device in the home network, greatly enriching the data source. The data analysis module uses a deep neural network and reinforcement learning algorithm to fuse and analyze multi-source data. Through the dynamic threshold adaptive algorithm and trend prediction model, the threshold for judging the operating status of the device can be automatically adjusted, so that the threshold accuracy is effectively improved. At the same time, the trend of network traffic changes can be accurately predicted, thereby achieving accurate judgment of the device operating status and early optimization of network performance, significantly enhancing the stability and reliability of the network and improving the user's network usage experience.

[0069] By building a dynamic collaborative decision-making module consisting of multiple sub-decision-making units, the system dynamically activates or deactivates corresponding sub-decision-making units based on different network scenario requirements, reducing system decision-making power consumption. In complex network environments or fault scenarios, it generates optimized control instructions based on abnormal signals, network traffic trends, and dynamically updated preset control policies. The execution of control instructions is monitored across multiple dimensions, including performance metrics, user experience, and energy consumption, and the execution process is documented using blockchain technology. This multi-dimensional comprehensive assessment and tamper-proof evidence storage mechanism not only provides a reliable traceability basis for the control instruction execution process, ensuring system security and auditability, but also facilitates a comprehensive and accurate assessment of system performance, providing richer and more effective feedback to the data analysis module, supporting continuous system optimization, and further enhancing the intelligent management level and overall performance of home gateway devices.

Claims

1. A control system for an intelligent PON home gateway device, characterized in that: It includes a data acquisition module, a data analysis module, a dynamic collaborative decision-making module, an adaptive execution module, and a feedback evaluation module. The data acquisition module is connected to the data analysis module, the data analysis module is connected to the dynamic collaborative decision-making module, the dynamic collaborative decision-making module is connected to the adaptive execution module, the adaptive execution module is connected to the feedback evaluation module, and the feedback evaluation module is connected to the dynamic collaborative decision-making module; The data acquisition module is used to collect the operating data of each module of the PON home gateway device and the application behavior data of each terminal device in the home network in real time; The data analysis module is used to perform a fusion analysis of the collected operation data of the gateway device and the application behavior data of the terminal device based on a deep neural network and a reinforcement learning algorithm; The dynamic collaborative decision-making module is used to build a decision-making network including multiple sub-decision-making units, and generate corresponding control instructions based on the real-time collected gateway device operation data and terminal device application behavior data; The adaptive execution module is used to receive the control instruction and intelligently assign the instruction execution priority according to the real-time load of each functional module of the home gateway device, transmit the instruction to the corresponding functional module, and drive the functional module to execute the instruction operation; The feedback evaluation module is used to monitor the execution effect of control instructions from multiple dimensions such as performance indicators, user experience, and energy consumption, and uses blockchain technology to store the execution process. The post-execution equipment operation data and evaluation results are fed back to the data analysis module for analysis. The dynamic collaborative decision-making module dynamically adjusts decisions based on the analysis results.

2. The control system of a smart PON home gateway device according to claim 1, characterized in that: The data acquisition module includes an edge computing node and a distributed sensor array. The distributed sensor array is deployed at designated key locations of the home gateway device. The edge computing node establishes a connection with each sensor through a communication protocol such as I2C and SPI. The distributed sensor array converts the collected physical signals into electrical signals and transmits them to the corresponding edge computing node. The node's built-in analog-to-digital converter converts the electrical signals into digital data. At the same time, the edge node communicates with the terminal devices in the home network through the network interface, using the simple network management protocol or the API interface customized by the device manufacturer to collect the application behavior data of the terminal devices. The collected data is preliminarily preprocessed at the edge computing node and packaged according to the preset data structure, and transmitted to the data analysis module through the high-speed data bus.

3. The control system of a smart PON home gateway device according to claim 2, characterized in that: The data analysis module includes a deep neural network processing unit and a reinforcement learning algorithm execution unit. After receiving various types of data obtained by each edge node, the deep neural network processing unit performs feature extraction on each type of data to obtain high-dimensional data features, and reduces the dimensionality of the high-dimensional data features through the principal component analysis algorithm to obtain low-dimensional feature vectors, and transmits the low-dimensional feature vectors to the reinforcement learning algorithm execution unit.

4. The control system of a smart PON home gateway device according to claim 3, characterized in that: The reinforcement learning algorithm execution unit constructs the environment state based on the low-dimensional feature vector, calculates the action value of each possible action, i.e., the Q value, based on the deep neural network processing unit, and selects the action based on the ∈-greedy strategy; the action that the current strategy network considers to have the largest Q value is selected with a probability of 1-∈.

5. The control system of a smart PON home gateway device according to claim 3, characterized in that: The execution unit of the reinforcement learning algorithm execution unit executes action A t After that, the state of the home gateway device will transition to the new state S t+1 The execution unit quantitatively evaluates the effect of the action based on the pre-designed reward mechanism and calculates the corresponding immediate reward R t+1 The reward function is designed based on the optimization goal specified by the anchoring system. The specified optimization goal includes network performance optimization or device stability optimization. In the reward calculation process, the reward attenuation factor γ is set, γ∈(0,1) to balance short-term rewards and long-term rewards, so that the algorithm pays more attention to long-term benefits. Gt=R t+1 +γR t+2 +γ2R t+3 +…; After executing action At, the actual operating state of the home gateway device is transferred from St to S t+1 The transfer process is determined by the actual operation mechanism of the home gateway device, including changing the signal strength and channel occupancy of the wireless communication module by adjusting the wireless channel parameters, which is reflected in the new state. t+1 In the process of executing the unit, the state transition is recorded, i.e., (St, At, Rt+1, St+1), and the data is stored in the experience replay buffer.

6. The control system of a smart PON home gateway device according to claim 1, characterized in that: The multiple sub-decision-making units of the dynamic collaborative decision-making module include an optical signal processing sub-decision-making unit, a CPU load management sub-decision-making unit, and a network traffic scheduling sub-decision-making unit. After receiving the analysis results of the data analysis module, the dynamic collaborative decision-making module first determines the activated sub-decision-making unit based on the current network scenario requirements through the attention mechanism. When the optical signal strength of the PON interface module is detected to be abnormal, the optical signal processing sub-decision-making unit is activated; if it is predicted that the network traffic will increase significantly, the network traffic scheduling sub-decision-making unit is activated at the same time; The sub-decision-making unit makes local decisions based on the preset control strategy and the information received. The preset control strategy is optimized and adjusted according to historical decision-making results and changes in the network environment through a reinforcement learning algorithm. When the optical signal strength is judged to be lower than the threshold, the Q-Learning algorithm is used to select a joint strategy of adjusting the optical module transmission power and switching to the backup optical path, combined with the optical signal recovery situation under different adjustment strategies in historical data, and generate corresponding control instructions. The decision results of multiple sub-decision-making units are integrated through a multi-objective optimization algorithm to balance the conflicts between different decisions.

7. The control system of a smart PON home gateway device according to claim 1, characterized in that: The adaptive execution module includes a software-defined network architecture, which includes a controller and a forwarder. After receiving the control instruction generated by the dynamic collaborative decision-making module, the controller first parses the instruction content to determine the functional modules and operation types of the home gateway device involved; then, through the real-time communication connection established with each functional module, obtains the current load status of the functional module; According to the load situation, the controller adopts a load balancing algorithm and assigns execution priority to control instructions through the minimum connection number algorithm. Urgent instructions that have a greater impact on the system are given a high priority; non-urgent optimization instructions are given a lower priority. After assigning the priority, the controller accurately transmits the instructions to the forwarder through the specified communication protocol. After receiving the instructions, the functional module executes the operation according to the instruction requirements and feeds back the execution results to the controller.

8. The control system of a smart PON home gateway device according to claim 1, characterized in that: The feedback evaluation module monitors and evaluates the execution effect of the control instructions from multiple dimensions such as performance indicators, user experience, and energy consumption; In terms of performance indicator monitoring, probes deployed at key network nodes collect data on network bandwidth utilization, latency, and packet loss in real time; Use the device's built-in performance monitoring module to obtain device processing capacity and response time indicators; In user experience evaluation, a combination of active detection and passive monitoring is used. Test data packets are sent to terminal devices in the home network to evaluate network quality. User feedback on network usage experience through the app or web page is passively collected. For energy consumption monitoring, smart meters are used to record the energy consumption information of home gateway devices in real time; Integrating blockchain technology, key data in the control instruction execution process is packaged according to the blockchain's block structure, and a unique block identifier is generated through a hash algorithm. Each block is linked in sequence to form an unalterable execution process evidence chain. The collected post-execution equipment operation data and evaluation results are integrated and analyzed to generate a feedback report which is transmitted to the data analysis module to provide a basis for subsequent data analysis and decision adjustments.

9. The control method of a smart PON home gateway device according to claim 1, characterized in that: The following steps are involved: S1: Real-time collection of operating data of each module of the PON home gateway device and application behavior data of each terminal device in the home network; S2: Based on deep neural networks and reinforcement learning algorithms, it integrates and analyzes the collected operation data of gateway devices and application behavior data of terminal devices; S3: Build a decision-making network consisting of multiple sub-decision-making units to generate corresponding control instructions based on the real-time collected gateway device operation data and terminal device application behavior data; S4: Based on the received control instruction and according to the real-time load of each functional module of the home gateway device, intelligently assigning the instruction execution priority, transmitting the instruction to the corresponding functional module, and driving the functional module to execute the instruction operation; S5: Monitor the execution effect of control instructions from multiple dimensions such as performance indicators, user experience, and energy consumption, use blockchain technology to store the execution process, and feed back the equipment operation data and evaluation results after execution to the data analysis module for analysis. The dynamic collaborative decision-making module dynamically adjusts decisions based on the analysis results.

Images